December 13 2002
Noise to the curvature distribution caused by surface atom identification
error
The curvature distribution at 1500K is invariant to the system size, but
it is sensitive to the identification of the surface atoms. That is, if
some of the internal atoms are regarded as the surface atoms, the curvature
distribution at 1500K is quite different.
N=2624 with different cutoffs of neighoring atoms
Different size with cutoff=3.5
Different size with cutoff=3.7
Curve fitting of Curvature distribution for 1409 atoms
Better data for exact surface atom discrimination are expected, but we
can first try to fit the on-hand data and get some qualitive impressions
about them.
The fitting fuction is the sum of two Gamma functions. The tails are
not fitted very well for T<=500K. Three Gamma functions may be the best.
The problem is that requires 12 variables but XMGR can only do fitting
with up to 10 variables. In the future work I may have to find another
tool or do it by c.
y = a0 * exp(-a1*x) * x^a2 + a3*exp(-a4*(x-a5))*x^a6+a7
The following pictures are those with 1409 atoms and have 1000 bins.